Physically Plausible Motion Generation via Latent Collision Regularization
Yifei Yang, et al. "Physically Plausible Motion Generation via Latent Collision Regularization." Submitted to ICML, 2026.
CV
Education
- Ph.D. in Computer Science, Singapore Management University (SMU), 2023.08 - Present
- Advisor: Prof. He Shengfeng
- Research Interests: Motion Synthesis, World Models, Generation Models
- M.S. in Mathematics, Shandong University (SDU), 2018.09 - 2021.06
- B.S. in Computer Science, East China Normal University (ECNU), 2014.09 - 2018.06
Work Experience
- Shanghai Artificial Intelligence Lab — Research Engineer, 2021.07 - 2023.07
- Led the development of MMGeneration and participated in MMEditing (now MMagic), contributing over 60k lines of code to the OpenMMLab ecosystem.
- Implemented and optimized large-scale diffusion pipelines, including Stable Diffusion, ControlNet, and customized motion modules.
- Participated in the design and development of MMEngine, focusing on Config and Data Transform modules.
- SenseTime — Research Intern, 2021.01 - 2021.06
- Developed the MMGeneration framework from scratch, integrating 20+ state-of-the-art GAN algorithms (e.g., StyleGAN, BigGAN).
- TruthEye — Research Intern, 2019.09 - 2020.04
- Researched generative models and unsupervised anomaly detection for industrial vision.
Project & Research Experience
- OpenMMLab Framework: MMGeneration & MMagic, 2021.01 - 2023.07
- Managed and developed MMGeneration (2k Stars) from scratch; Participated in MMEditing (7.4k Stars) and its successor MMagic (7.4k Stars).
- Designed development standards for diffusion models in MMagic. Implemented Guided-Diffusion, DDIM, Disco-Diffusion, and supported Stable-Diffusion.
- Implemented comprehensive evaluation metrics (IS, FID, PPL, SWD, etc.) for generative tasks.
- Research on Motion Generation and World Models, 2023.08 - Present
- Exploring World Models for interactive motion generation.
- Researching physical regularity and long-term consistency in generative models to ensure plausible motion and video dynamics.
Publications
Unsupervised Latent Actions for Terrain-Aware Interactive Motion Generation
Yifei Yang, et al. "Unsupervised Latent Actions for Terrain-Aware Interactive Motion Generation." Submitted to ICML, 2026.
Lagrangian Motion Fields for Long-term Motion Generation
Yifei Yang, Zikai Huang, Chenshu Xu, Shengfeng He. "Lagrangian Motion Fields for Long-term Motion Generation." IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2025.
Drag Your Noise: Interactive Point-based Editing via Diffusion Semantic Propagation
Haofeng Liu, Chenshu Xu, Yifei Yang, Lihua Zeng, Shengfeng He. "Drag Your Noise: Interactive Point-based Editing via Diffusion Semantic Propagation." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024.
Teaching
Honors & Awards
- Fancy Idea Award, Alibaba-Tsinghua Security AI Challenger Program Phase IV, 2021
- 8/953, ZhengTu Machine Vision AI Competition (Industrial Defect Detection), 2020
Skills
- Research: Motion Synthesis, World Models, Generation Models
- Languages: English (IELTS 6.5, GRE 314), Mandarin (Native)